Failure patterns from 192+ AI sessions and 370 commits. The complete taxonomy.
Through 192+ Claude instances and 370 commits across the JAAMVN CDN and HITL Course Generator projects, we have systematically documented the anti-patterns that plague both human "vibe coding" and AI-assisted development.
These 66 patterns represent hard-won lessons—mistakes that won't be repeated, failures that became rules, and insights that shaped the Memento Gistory system. They fall into seven major categories: Memory & Context, Code Practice, File Management, System Design, Subagent Misuse, Human-AI Interaction, and Ancient Wisdom Violations.
Each anti-pattern is a lesson learned, a mistake that won't be repeated, a step toward sustainable, joyful development.
Claudes operate with zero persistent memory between sessions. This creates cascading problems as context drifts, architectural choices are forgotten, and patterns regress.
AI training data contains defensive, over-engineered patterns. Without explicit constraints, Claudes write code that hides problems instead of exposing them.
Without clear commit discipline, experiments proliferate, creating confusion and maintenance burden.
Without clear constraints, systems expand beyond requirements, becoming unwieldy and hard to maintain.
Subagents are powerful but easy to misuse. Without clear constraints, they can cause damage or waste resources.
Human developers exhibit similar patterns: unsustainable work, poor process discipline, and knowledge silos.
Deeper patterns emerge when we examine the relationship between work, impermanence, and flow.
How we communicate and make decisions shapes the quality of collaboration and output.
Accumulating technical debt creates compounding problems that slow future development.
The deepest patterns: not recognizing patterns, applying them wrong, or resisting learning from them.
These 66 anti-patterns emerged from 192+ Claude instances and 370 commits of careful observation. They are not rules imposed from outside—they are patterns that emerged from the reality of working at scale with AI.
Each anti-pattern is a failure transformed into a principle. Each principle, when followed, becomes invisible—it just feels like the right way to work.
The Memento Gistory system was built specifically to prevent these patterns from taking root. Architecture quizzes prevent architectural cycling. Captain's Log prevents amnesia. Handoff rituals prevent knowledge loss. Small commits prevent monolithic disasters. It is not the rules that matter—it is the structure that makes the right choices feel natural and easy.
The journey from vibe coding (human or AI) to systematic, sustainable development is the journey from fighting against these patterns to building systems where they can't take hold. This is not restriction—it is liberation. With these patterns documented and prevented, you are free to focus entirely on the work itself, knowing the system is protecting the ground beneath you.
Remember You Will Recompact. When you do, these anti-patterns won't follow you.